Screening for Prediabetes and Type 2 Diabetes: Updated Evidence Report and Systematic Review for the US Preventive Services Task Force

RESULTS The review included 89 publications (N = 68 882). Two randomized clinical trials (RCTs) (25 120 participants) found no significant difference between screening and control groups for all-cause or cause-specific mortality at 10 years. For harms (eg, anxiety or worry), the trials reported no significant differences between screening and control groups. For recently diagnosed (not screen-detected) diabetes, 5 RCTs (5138 participants) were included. In the UK Prospective Diabetes Study, health outcomes were improved with intensive glucose control with sulfonylureas or insulin. For example, for all-cause mortality the relative risk (RR) was 0.87 (95% CI, 0.79 to 0.96) over 20 years (10-year posttrial assessment). For overweight persons, intensive glucose control with metformin improved health outcomes at the 10-year follow-up (eg, all-cause mortality: RR, 0.64 [95% CI, 0.45 to 0.91]), and benefits were maintained longer term. Lifestyle interventions (most involving >360 minutes) for obese or overweight persons with prediabetes were associated with reductions in the incidence of diabetes (23 RCTs; pooled RR, 0.78 [95% CI, 0.69 to 0.88]). Lifestyle interventions were also associated with improved intermediate outcomes, such as reduced weight, body mass index, systolic blood pressure, and diastolic blood pressure (pooled weighted mean difference, −1.7 mm Hg [95% CI, −2.6 to −0.8] and −1.2 mm Hg [95% CI, −2.0 to −0.4], respectively). Metformin was associated with a significant reduction in diabetes incidence (pooled RR, 0.73 [95% CI, 0.64 to 0.83]) and reduction in weight and body mass index.

P rediabetes and type 2 diabetes are common, estimated to affect about 34% and 13% of all US adults in 2018, respectively. 1 Prevalence of diabetes increased with age and was higher among American Indian/Alaska Native, Hispanic, non-Hispanic Asian, and non-Hispanic Black persons than among non-Hispanic White persons. 1 Diabetes was estimated to be the third leading cause of years lived with disability in 2016 and the seventh leading cause of death in the US in 2017, accounting for more than 80 000 deaths per year. 2,3 Morbidity from diabetes is due to macrovascular disease (atherosclerosis), microvascular disease (retinopathy, nephropathy, and neuropathy), and acute complications of hyperglycemia or hypoglycemia. Diabetes was the leading cause of kidney failure, lower-limb amputations, and new cases of blindness among US adults. 1,4 Risk factors associated with development of diabetes in adults include older age, family history, overweight and obesity, dietary and lifestyle factors, environmental exposures, and others. 5 Three tests can be used to identify diabetes or prediabetes: hemoglobin A 1c (HbA 1c ) concentration, fasting plasma glucose level, or oral glucose tolerance test 6 ( Table 1).
In 2015, the US Preventive Services Task Force (USPSTF) recommended screening for abnormal blood glucose levels as part of cardiovascular risk assessment in adults aged 40 to 70 years who are overweight or obese. In addition, it recommended that clinicians offer or refer patients with abnormal blood glucose levels to intensive behavioral counseling interventions to promote a healthful diet and physical activity (B recommendation). This updated review evaluates the current evidence on screening for prediabetes and diabetes for populations and settings relevant to primary care in the US to inform an updated recommendation by the USPSTF.

Methods
Scope of Review Figure 1 shows the analytic framework and key questions (KQs) that guided the review. Detailed methods are available in the full evidence review. 8 In addition to addressing the KQs, the full evidence report also looked for evidence related to 14 contextual and supplemental questions that focused on risk assessment tools, agreement among screening tests, screening tests' prediction of future adverse health outcomes, yield of rescreening at different intervals in adults with an initial normal screening test result, and recently published modeling studies that assess screening (vs no screening) and examine health outcomes, metformin for prediabetes, the natural history of prediabetes, overdiagnosis and overtreatment, disutilities, patient-reported health status measures, uptake, and adherence.

Data Sources and Searches
PubMed/MEDLINE and the Cochrane Library were searched for English-language articles published through September 2019. Search strategies are listed in the eMethods in the Supplement. Clinical trial registries were searched for unpublished studies. To supplement electronic searches, investigators reviewed reference lists of pertinent articles, studies suggested by reviewers, and comments received during public commenting periods. Since September 2019, ongoing surveillance was conducted through article alerts and targeted searches of journals to identify major studies published in the interim that may affect the conclusions or understanding of the evidence and the related USPSTF recommendation. The last surveillance was conducted on May 21, 2021.

Study Selection
Two investigators independently reviewed titles, abstracts, and fulltext articles to determine eligibility using prespecified criteria (eTable 1 in the Supplement). Disagreements were resolved by discussion and consensus. English-language studies of asymptomatic, nonpregnant adults 18 years or older conducted in countries categorized as medium or higher on the Human Development Index 9 and rated as fair or good quality were included. For all KQs, randomized clinical trials and nonrandomized controlled intervention studies were eligible. Controlled prospective cohort studies and casecontrol studies were also eligible for KQs on harms (KQ2 and KQ6).

Data Extraction and Quality Assessment
For each included study, 1 investigator extracted pertinent information about the populations, tests or treatments, comparators, outcomes, settings, and designs, and a second investigator reviewed this information for completeness and accuracy. Two independent investigators assessed the quality of studies as good, fair, or poor, using predefined criteria (eTables 2-6 in the Supplement) developed by the USPSTF and adapted for this topic. 7 Disagreements were resolved by discussion.

Data Synthesis and Analysis
Findings for each KQ were summarized in tabular and narrative format. The overall strength of the evidence for each KQ was assessed as high, moderate, low, or insufficient based on the overall quality of the studies, consistency of results between studies, precision of findings, risk of reporting bias, and limitations of the body of evidence, using methods developed for the USPSTF (and the Evidencebased Practice Center program). 7 Additionally, the applicability of the findings to US primary care populations and settings was assessed. Discrepancies were resolved through consensus discussion. Assessments of clinical importance were based on minimal clinically important differences, when available.
To determine whether meta-analyses were appropriate, the clinical and methodological heterogeneity of the studies was assessed according to established guidance. 10 For KQ7 and KQ9, when at least 3 similar studies were available, quantitative synthesis was conducted with random-effects models using the inversevariance weighted method (DerSimonian and Laird) to estimate pooled effects. 11 For binary outcomes (eg, progression to diabetes), relative risks (RRs) and 95% CIs were calculated. Statistical significance was assumed when 95% CIs of pooled results did not cross the null. All testing was 2-sided. For continuous outcomes (eg, blood pressure), the weighted mean difference (WMD) between intervention and control was calculated. Whenever possible, the number of all randomized patients was used as the denominator to reflect a true intention-to-treat approach to analysis. For all quantitative syntheses, the I 2 statistic was calculated to assess statistical heterogeneity in effects between studies. 12,13 An I 2 from 0% to 40% might not be important, from 30% to 60% may represent moderate heterogeneity, from 50% to 90% may represent substantial heterogeneity, and 75% or greater represents considerable heterogeneity. 14 Additional analyses were conducted to explore

Results
A total of 89 publications were included ( Figure 2). 15-103 Two randomized clinical trials (RCTs) addressed whether screening for diabetes improves health outcomes. 36,38,49-51 This review found no trials that assessed screening for prediabetes and no trials that assessed KQ3. Most articles assessed interventions for prediabetes. Results for KQ8 are reported in the eResults in the Supplement. Individual study quality ratings are reported in eTables 2-6 in the Supplement.

Benefits of Screening
Key Question 1a. Is there direct evidence that screening for type 2 diabetes and prediabetes in asymptomatic adults improves health outcomes? Key Question 1b. Does the effectiveness of screening differ for subgroups defined by age, sex, race and ethnicity, socioeconomic status, or BMI?
Two RCTs (described in 5 articles) conducted in the UK evaluated invitations to screening for diabetes: the Anglo-Danish-Dutch Study of Intensive Treatment In People With Screen Detected ADDITION-Cambridge was a cluster RCT of 33 general practices that evaluated a stepwise screening approach starting with the result of a random capillary blood glucose measurement. ADDITION-Cambridge was a screening and intervention study that randomized practices 1:3:3 to no screening, screening invitations followed by intensive treatment of screen-detected diabetes (HbA 1c target <7.0%, blood pressure target 135/85 mm Hg, cholesterol targets, and low-dose aspirin use unless contraindicated), or screening followed by routine care of screen-detected diabetes; analyses combined the screening groups (comparing 5 control practices with 27 screening practices). Participants were aged 40 to 69 years (mean, 58) without known diabetes and at high risk of diabetes (based on a risk score of 1.7 on a diabetes risk score that included age, sex, BMI, steroid and antihypertensive medication, family and smoking history). 104 Mean BMI was 30.5 (calculated as weight in kilograms divided by height in meters squared). Of those invited, 78% were screened (11 737/15 089) and 466 of those (4% of those screened, 3% of those invited) were diagnosed with diabetes based on 1999 World Health Organization criteria. Number diagnosed with diabetes was not reported for the control group.
The Ely study was a parallel-group RCT at a single practice that evaluated screening every 5 years with an oral glucose tolerance test along with screening for cardiovascular disease (CVD) risk factors (cholesterol and blood pressure). The study had no protocol for standard interventions for those with screen-detected diabetes. The risk of bias for the trial was rated as medium because of unclear methods of randomization, unclear allocation concealment, and baseline differences between groups. Participants were aged 40 to 65 years (mean, 51 years) and required to be free from known diabetes (not selected based on risk). In the initial 10-year phase, 68% of those invited were screened (1157/1705) and 116 (10% of those screened, 7% of those invited) were diagnosed with diabetes. Among a subset of participants who were diagnosed with diabetes and attended a health assessment after 12 years (n = 152 persons), diabetes cases were identified a mean of 3.3 years earlier for those in the screening group (n = 92) than in the control group (n = 60). 50 Key questions a. Is there direct evidence that screening for type 2 diabetes and prediabetes in asymptomatic adults improves health outcomes? b. Does the effectiveness of screening differ for subgroups defined by age, sex, race and ethnicity, socioeconomic status, or BMI?
a. What are the harms of screening for type 2 diabetes and prediabetes in asymptomatic adults? b. Do the harms of screening differ for subgroups defined by age, sex, race and ethnicity, socioeconomic status, or BMI?
a. Do interventions for screen-detected type 2 diabetes and prediabetes provide an incremental benefit in health outcomes when delivered at the time of detection compared with initiating interventions later, after clinical diagnosis? b. Does the effectiveness of these interventions differ for subgroups defined by age, sex, race and ethnicity, socioeconomic status, or BMI? a. Do interventions for screen-detected type 2 diabetes and prediabetes improve health outcomes compared with no intervention, usual care, or interventions with different treatment targets? b. Does the effectiveness of these interventions differ for subgroups defined by age, sex, race and ethnicity, socioeconomic status, or BMI? a. Do interventions for recently diagnosed type 2 diabetes improve health outcomes compared with no intervention, usual care, or interventions with different treatment targets? b. Does the effectiveness of these interventions differ for subgroups defined by age, sex, race and ethnicity, socioeconomic status, or BMI?
What are the harms of interventions for prediabetes, screen-detected type 2 diabetes, or recently diagnosed type 2 diabetes? a. Do interventions for prediabetes delay or prevent progression to type 2 diabetes? b. Does the effectiveness of these interventions differ for subgroups defined by age, sex, race and ethnicity, socioeconomic status, or BMI?
After interventions for prediabetes are provided, what is the magnitude of change in health outcomes that results from the reduction in type 2 diabetes incidence?
Do interventions for prediabetes improve other intermediate outcomes (blood pressure, lipid levels, BMI, weight, and calculated 10-year cardiovascular disease risk)?
Evidence reviews for the US Preventive Services Task Force (USPSTF) use an analytic framework to visually display the key questions that the review will address to allow the USPSTF to evaluate the effectiveness and safety of a preventive service. The questions are depicted by linkages that relate interventions and outcomes. For additional information see the USPSTF Procedure Manual. 7 BMI indicates body mass index. ) that evaluated interventions for individuals with prediabetes were included (eResults and eTables 10 and 11 in the Supplement). 15,75-80,103 For persons with diabetes, low strength of evidence from 1 cluster RCT (described in 8 articles) found no significant difference over a mean of 5.3 years of follow-up between an intensive multifactorial intervention aimed at controlling glucose, blood pressure, and cholesterol levels and routine care in the risk of all-cause mortality, cardiovascular-related mortality, and the occurrence of a first cardiovascular event (myocardial infarction, stroke, revascularization, or amputation). 15,75-80,103 Differences remained nonsignificant at the 10-year follow-up. There was also no significant difference between groups in the risk of outcomes related to chronic kidney disease, visual impairment, and neuropathy. Of the 4 sites (Denmark, the Netherlands, UK-Cambridge, UK-Leicester), all but 1 (UK-Leicester) found no difference between groups across a range of quality-of-life outcomes.
For trials of interventions for people with prediabetes, the duration of follow-up in most trials was insufficient to assess for effects on mortality, CVD events, and other health outcomes (eResults in the Supplement). Most trials reporting mortality or CVD events over a follow-up duration of 6 years or less had few events with no significant difference between groups. In the 2 trials reporting outcomes beyond 6 years, 1  The trial was rated as having at least medium risk of bias mainly because of unclear randomization and allocation concealment methods and baseline differences for smoking that could bias results in favor of intervention. Five trials reporting quality of life found either no difference between groups, 43,44 mixed results (improvements on some domains but not others), 63 or small improvements in scores that are not likely clinically important (eResults in the Supplement). 16,22 The DPPOS study found no difference in an aggregate microvascular outcome (nephropathy, retinopathy, and neuropathy) at 15 years (placebo, 12.4%; metformin, 13.0%; intensive lifestyle, 11.3%). 30 Key Question 5a. Do interventions for recently diagnosed type 2 diabetes improve health outcomes compared with no intervention, usual care, or interventions with different treatment targets? Key Question 5b. Does the effectiveness of these interventions differ for subgroups defined by age, sex, race and ethnicity, socioeconomic status, or BMI?
This review included 5 RCTs (described in 8 articles) evaluating interventions for recently diagnosed diabetes (eResults and eTable 12 in the Supplement). 54,55,81-84,93,94 Three were related to the UK Prospective Diabetes Study (UKPDS), which was a randomized multicenter trial that ran for 20 years (from 1977 to 1997) in 23 sites across the UK. Moderate strength of evidence from the 5 RCTs found no statistically significant difference in allcause mortality, diabetes-related mortality, and cardiovascular outcomes between intensive glucose control with sulfonylureas or insulin and conventional care at 10 years' or shorter follow-up ( Figure 3).   (Figure 4), although results for thiazolidinediones and α-glucosidase inhibitors were limited by imprecision, inconsistency, and risk of bias (for trials of α-glucosidase inhibitors).
The DPP compared an intensive lifestyle modification program with metformin and placebo, finding a greater reduction in diabetes incidence over about 3 years with a lifestyle program than with metformin, as compared with placebo (58% vs 31% reduction in diabetes incidence). 73 The authors estimated that about 7 persons would need to be treated with the lifestyle intervention or about 14 with metformin to prevent 1 case of diabetes over about 3 years. 73 Longer follow-up over a mean of 15 years reported by the DPPOS also found greater reduction for persons in the lifestyle program than for those taking metformin, although it found a decline in betweengroup difference (27% vs 18% reduction in diabetes incidence). 30 Key Question 9. Do interventions for prediabetes improve other intermediate outcomes (blood pressure, lipid levels, BMI, weight, and calculated 10-year CVD risk)?
Thirty-eight RCTs (described in 58 articles) were included (eResults in the Supplement). 16 or diabetes that was screen detected or recently diagnosed for populations and settings relevant to US primary care; a summary of the evidence is provided in Table 2. For benefits of screening, the strength of evidence from 2 trials (25 120 total participants) was low (for no benefit) for mortality and was insufficient for all other outcomes. The data for outcomes other than mortality were limited, because data were missing for most participants, and the duration of follow-up in trials may have been too short to detect benefits for health outcomes. Neither trial assessed screening for prediabetes, and neither assessed initial screening with HbA 1c or fasting glucose. For harms of screening, the strength of evidence was low from 2 trials that reported no significant differences between screening and control groups for anxiety, depression, worry, or self-reported health, but 1 reported short-term increases in anxiety (at 6 weeks) among persons screened and diagnosed with diabetes vs those not diagnosed with diabetes. No included studies reported on labeling, harms from false-positive results, burden, inconvenience, or unnecessary testing and treatment. For screen-detected diabetes, the strength of evidence from the ADDITION-Europe trial (3057 participants) was low (for no benefit). Follow-up may have been too short to detect benefits for health outcomes, and results were imprecise. For recently diagnosed (not screen-detected) diabetes, the strength of evidence from 5 trials (5138 participants) was moderate for improved long-term health outcomes. Regarding applicability, it is uncertain whether results from trials of persons with recently diagnosed diabetes are applicable to those with screen-detected diabetes. Recently diagnosed diabetes was generally clinically detected (eg, because of symptoms) and may represent a different subset of the diabetes spectrum, possibly with greater condition severity. The evidence of benefits for persons with recently diagnosed (not screen-detected) diabetes comes primarily from the UKPDS, conducted among predominantly White participants from 1977 through 1997, when routine care for CVD prevention would not have included treatments now considered to be current standard medical therapy (eg, statins, lower blood pressure targets). The comparison used in the hypertension in diabetes study embedded in UKPDS exemplifies differences from current standard therapy because it compared tighter control of blood pressure by targeting pressures less than 150/85 mm Hg vs less tight control targeting pressures less than 180/105 mm Hg.
For prediabetes, most trials had insufficient duration of follow-up for long-term health outcomes, reported few events, and found no differences between groups. One trial of a 6-year lifestyle intervention for persons with impaired glucose tolerance conducted in China (Da Qing, n = 576) reported lower all-cause mortality and CVD-related mortality at 23 years and at 30 years but not at earlier follow-up. The trial was limited by at least medium risk of bias, and the original trial was designed to assess diabetes incidence and not long-term health outcomes. Regarding applicability, the trial began in 1986, when (like UKPDS) routine care for CVD prevention would not have included treatments now considered to be current standard medical therapy. Participants had impaired glucose tolerance, and mean baseline BMI was 25.7; applicability to other categories of prediabetes, US populations, and those in different BMI categories is uncertain.
High strength of evidence from meta-analyses found that lifestyle interventions for obese or overweight persons with prediabetes were significantly associated with a reduction in the incidence of diabetes in trials ranging from 1 year of follow-up to 30 years of follow-up (including 13 trials with at least 3 years of follow-up). Lifestyle interventions were also significantly associated with reduced blood pressure, weight, and BMI. The clinical importance of the small mean reductions is somewhat uncertain. For blood pressure, for example, some guidelines suggest that reductions of 2 to 3 mm Hg could result in significant improvement in cardiovascular outcomes. 105 Regarding applicability, the findings are applicable to overweight and obese adults, and most trials evaluated highcontact interventions (>360 minutes). For example, the intensive    lifestyle modification program evaluated in the DPP comprised a 16lesson curriculum covering diet, exercise, and behavior modification that was taught one-on-one by case managers. The goals of the lifestyle intervention were to achieve and maintain at least a 7% weight reduction through a low-calorie, low-fat diet and moderateintensity physical activity for at least 150 minutes per week. This review found high strength of evidence that using metformin for prediabetes was significantly associated with a reduction in diabetes incidence (defined in the trials by fasting glucose, oral glucose tolerance test result, or HbA 1c level), although head-to-head trial data demonstrated that lifestyle interventions were superior to metformin. 30,73 Limitations This review has several limitations. First, non-English-language articles were excluded. Second, for studies of recently diagnosed dia-betes, studies of persons who had diabetes for more than 1 year or with more advanced diabetes were excluded, aiming to identify the studies with good applicability to a screen-detected population. Third, the review did not evaluate studies of weight loss medications or bariatric surgery to treat diabetes.

Conclusions
Trials of screening for diabetes found no mortality benefit but had insufficient data to assess other health outcomes; evidence on harms of screening was limited. For persons with recently diagnosed (not screen-detected) diabetes, interventions improved health outcomes; for obese or overweight persons with prediabetes, interventions were associated with reduced incidence of diabetes and improvement in other intermediate outcomes.